HronTurekFSI3: Force calculation for the fsi tutorial, problems with validation
 

Dear all,

I am trying to perform a validation of the solids4Foam based on the HronTurekFSI3 benchmark. I started the case basing on the tutorial added to the solids4Foam. First of all, I noticed that the force function in the controldict file has an incorrect value of rhoInf; I guess it should be 1000? and calcualtes force solely over the plate, not the plate + cylinder.

More importantly the force.gnuplot file, designed to plot the forces divide the force value over 0.015, and in that manner we reach the expected values of lift and drag. The units are set to [N]. Why we need to divide the forces over 0.015? Is it related to the solids4foam? Maybe somebody has already encounter the same problem?

I am enclosing a file with the results for the drag calculation for the fluid part of the benchmark (CFD3). The solids4Foam solver gives comparable results only when I divide over 0.015. Is the 0.015 realted to the area of the cylinder and the plate?

https://i.imgur.com/n3OEodO.png

I tried to set to calculate the force coefficients instead, but I get information:

'Unknown function type forceCoeffs'

Is there any way to get force coefficients with use of solids4foam?

Here is the controlDict file part related to the force calculation. I added my comments in the parts where in my opinion needs to be updated.

plot [0.01:] "< sed s/[\\(\\)]//g ./forces/0/forces.dat" using 1:($2+$5)/0.015
forces
{
type forces;
functionObjectLibs ( "libforces.so" );
outputControl timeStep;
outputInterval 1;
patches (plate); / /(plate cylinder)?
pName p;
UName U;
rhoName rhoInf;
log true;
rhoInf 100; //should be 1000?
CofR (0.5 0.1 0);

region fluid;
}
);

Please let me know, if someone of you encounter similar doubts.

Best regards, Justyna

Hi,
I received an anwer to my problem from Philip Cardiff, here it is:

"Hi Justyna,

The reason for dividing by 0.015 Is because the domain is 0.015 m thick in the Z direction, so dividing by 0.015 gives us the force per unit depth i.e. for a depth of 1 m (as specified in the paper).

You then might wonder why was the model setup to use a thickness of 0.015 m since it is 2-D and it should not matter. I think the reason for this is twofold:
- The model would be awkward to manipulate in ParaView if the Z dimension was 1 m
- (I think, but I would need to double-check) Interface geometrical tolerances depend on the size of the interface faces; if we used a model with such long faces in the Z direction, it may result in artificially large tolerances, so it is better to choose a Z dimension similar to the X/Y dimensions. "

I hope that helps! Regards, Justyna